Accuracy and Precision of Continuous Noninvasive Arterial Pressure Monitoring Compared with Invasive Arterial Pressure: A Systematic Review and Meta-analysis

INTERMITTENT arterial pressure monitoring is part of the American Society of Anesthesiologists’ standards for all patients undergoing anesthesia.^* In patients undergoing high-risk surgery and/or presenting with major comorbidities, invasive arterial pressure monitoring is often used as the standard of care. This technique allows continuous, beat-to-beat arterial pressure monitoring as well as access for blood draws and is used in approximately 10 to 12% of all patients undergoing anesthesia in the United States and in Europe.¹  However, this invasive monitoring is associated with mechanical, infectious, and thrombotic complications.^2,3 

Recently, continuous noninvasive arterial pressure monitoring systems based on the volume clamp method and arterial tonometry have been developed and are now available at the bedside: Nexfin (BMEYE B.V., Amsterdam, The Netherlands); CNAP (CNSystems, Graz, Austria); and T-line (Tensys Medical, Inc., San Diego, CA).⁴  These devices display real-time, continuous arterial pressure waveforms and allow noninvasive beat-to-beat arterial pressure measurement. The main advantage of these technologies is that they can bridge the gap between noninvasive but intermittent oscillometric techniques and continuous but invasive arterial pressure monitoring. To date, these techniques have only been evaluated in small, single-center clinical studies, and no definitive validation studies have yet been performed.

We performed a systematic review and meta-analysis of studies that compared continuous noninvasive arterial pressure measurements with invasive arterial pressure measurements in adult patients in the perioperative and critical care settings. The principal outcomes were the accuracy and precision of continuous noninvasive systolic arterial pressure (SAP), diastolic arterial pressure (DAP), and mean arterial pressure (MAP) compared with invasive arterial pressure measurements. Accuracy and precision were defined as acceptable if bias was not greater than 5 mmHg and precision not greater than 8 mmHg.

This systematic review and meta-analysis was conducted following the guidelines set forth in Preferred Reporting Items for Systematic Reviews and Meta-Analyses.⁵ 

The following characteristics were defined in advance as eligibility criteria for the studies to be included in our systematic review and meta-analysis:

1. Published studies comparing arterial pressure measured using commercially available continuous noninvasive arterial pressure monitoring systems with that measured by invasive arterial pressure monitoring.

2. Studies reporting extractable bias and SD of the differences (or 95% limits of agreement [LOA]) between continuous noninvasive arterial pressure monitoring systems and invasive arterial pressure monitoring.

3. Studies on adult patient populations (age ≥18 yr) that report identifiable demographic data (sex and age).

4. Studies performed in the perioperative and critical care settings.

Three electronic databases (PubMed, Web of Science, and the Cochrane Library) were searched using the following key words: blood pressure, arterial pressure, monitor, measurement, noninvasive, Nexfin, CNAP, T-line, Finapres, Penaz, Vasotrac, volume clamp, pulse transit time, Wesseling, vascular unloading, preoperative, postoperative, perioperative, continuous, beat-to-beat, surgery, operative, anesthesia, intensive care unit, accuracy, precision, bias, limit of agreement, and Bland-Altman. The full electronic PubMed search strategy is presented in appendix 1. We restricted the search and subsequent bibliographic review to studies in the English language conducted on adult human subjects (≥18 yr old), and to published research articles (no case reports or correspondence). We also limited the search to studies expressing results as bias and either SD or 95% LOA. No restrictions were placed on the dates of the studies in our database search. In addition to the database search, we contacted the manufacturers of clinically available monitors—Nexfin, CNAP, and T-line—for other studies and hand-searched references in the studies included in the full-text retrieval for studies that had not been identified with the initial search.

Three investigators (S.-H.K., M.L., and K.S.S.) initially screened potentially eligible studies first by title and abstract. Remaining studies were then retrieved in full text. S.-H.K. and M.L. assessed eligibility according to inclusion criteria. If the eligibility of the study remained unclear, a third investigator (M.C.) made the final decision.

S.-H.K. and M.L. performed data extraction independently. A pilot data-extraction sheet was first used in five studies and then assessed for completeness and accuracy between S.-H.K. and M.L. Discrepancies between investigators were noted, the sheets updated, and then S.-H.K. and M.L. independently performed data extraction on the remaining studies. All data were then transferred separately to a standard Excel (Microsoft Corporation, Redmond, WA) spreadsheet. S.-H.K. and M.L. reviewed each other’s extractions for inconsistencies and, if needed, returned to the original work to validate the correct data.

Extracted study variables included patient age (mean, median, SD, range or interquartile range), sex, study setting (types of surgery, perioperative or critical care setting), number of patients enrolled in the study, and numbers of patients actually included in the analysis.

We extracted bias and SD of biases between invasive arterial pressure and the noninvasive arterial pressure measurement for SAP, DAP, and MAP from tables and Results sections of each article. If a study presented only bias and 95% LOA, SD was calculated as (upper LOA minus bias) divided by 1.96. As description of bias was not uniform among the studies (some articles described it as noninvasive minus invasive measurements, whereas others described it as invasive minus noninvasive measurements), we standardized bias in the current meta-analysis to mean noninvasive measurement minus invasive arterial pressure measurement and corrected source data as needed for reporting in this form. Authors of included studies were contacted to provide data if they were not published.

As there are no specific guidelines for the quality assessment of articles screened for inclusion in a meta-analysis focusing on method-comparison studies, we used modified Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) guidelines in order to meet our needs.⁶  Briefly, the original QUADAS-2 guidelines include four domains for the assessment of risk of bias (patient selection, index test, reference standard, and flow and timing) and three domains for the assessment of concerns related to applicability (patient selection, index test, and reference standard). Each domain consists of signaling questions that are marked yes, no, or unclear for the assessment of the quality of study; these are detailed in appendix 2. Three investigators (S.-H.K., M.L., and K.S.S.) modified these guidelines in order to make them suitable for the current meta-analysis. Then, three investigators (S.-H.K., M.L., and K.S.S.) performed an independent pilot assessment on a set of five articles. Quality indicators were compared, updated, and the pilot assessment repeated until all three investigators’ assessments became consistent. Then, two investigators (S.-H.K. and M.L.) performed independent quality assessments on the full study set using the final modified QUADAS-2 form (appendix 2). Risk for each of the bias domains and each of the applicability domains is classified as low, high, or unclear according to the modified QUADAS-2 guidelines. Disagreement between investigators was resolved by discussion with a fourth investigator (M.C.).

Principal summary measures of the current meta-analysis were (1) accuracy of measurement (defined as noninvasive − invasive measurement, or bias); (2) precision of measurement (defined as SDs of accuracy); and (3) 95% LOA of SAP, DAP, and MAP. The accuracy and precision were evaluated for SAP, DAP, and MAP and were a priori (in advance of the meta-analysis) defined as acceptable if accuracy was no greater than 5 mmHg and precision not greater than 8 mmHg for SAP and DAP, based on the standards for the validation of automatic arterial pressure monitoring established by the Association for the Advancement of Medical Instrumentation (AAMI).⁷  Of note, this definition of acceptability was developed by the AAMI for the evaluation of automatic sphygmomanometers. To the best of our knowledge, no official guidelines presently exist for the evaluation of continuous noninvasive arterial pressure monitoring systems, but the AAMI guidelines have been cited as a reference for acceptability in several recently published studies evaluating commercially available devices for continuous noninvasive arterial pressure measurement.^8–17  In addition, although most recently published studies define bias as the instantaneous absolute difference between noninvasive and invasive measurements, the AAMI guidelines define mean error as follows:

Consequently, bias as reported in method-comparison studies using Bland-Altman analysis^18,19  would result in a greater mean error and SD than the way these standards recommend for comparison of a new sphygmomanometer with invasive arterial pressure measurements.

For the synthesis of pooled estimates of bias and SD, we used random-effects models.^20,21  We tested heterogeneity of biases and SDs across studies using a Q test²¹  and quantified them with an I²  index.²²  The I²  statistic describes the percentage of variation across studies that is caused by heterogeneity rather than chance. If there were significant heterogeneity (I²  > 50%), we performed sensitivity analysis and meta-regression based on plausible clinical causes of the heterogeneity.²³  Forest plots are presented with individual and random-effects pooled estimates of bias and 95% LOA to visualize the data.

To assess for publication bias we created funnel plots for bias of SAP, DAP, and MAP against standard error for each study. These funnel plots were assessed visually for symmetry. In the absence of bias, these plots should resemble a symmetrical inverted funnel. To formally test for asymmetry, we applied Egger regression tests on secondary funnel plots using a significance level of 0.1 because of small sample size.²⁴ 

We conducted sensitivity and subgroup analyses to explore the causes of heterogeneity. These were performed based on the funding source (department vs. industry), identification of outliers (studies falling outside of mean ± 2SD for bias), setting (perioperative vs. critical care), current commercial availability (commercially available vs. unavailable devices), measurement site of invasive arterial pressure (radial vs. femoral), statistical approach (modified Bland-Altman analysis for repeated measurement vs. original Bland-Altman analysis), and risk of bias (low risk vs. high/unclear risk in flow and timing domain according to the modified QUADAS 2). We also conducted a meta-regression analysis on demographic characteristics (age, sex, and body mass index) and year of publication in addition to the study characteristics assessed in subgroup analysis. All the calculations and tests were conducted using Microsoft Excel 2010 (Microsoft Corporation) and R.^† Data are presented as mean ± SD or bias ± SDs (95% LOA).

As of May 8, 2013, a total of 574 articles were retrieved from the database searches and manufacturers after removing duplicates. Three investigators excluded 533 studies by title and abstract screening. The remaining 41 studies were retrieved as full-text articles and were assessed for eligibility. Thirteen articles^25–37  were excluded after full-text review for failure to meet the inclusion criteria or insufficient data for meta-analysis (appendix 3). The remaining 28 studies^{8–17,38–55} were included in the systematic review (fig. 1).

A total of 919 patients (65% male) were included in this meta-analysis. Most studies had small sample sizes (median, 25; range, 8 to 100). Among the 28 studies, 20^{9–12,14–17,40,41,43,44,46–49,52–55} were conducted in the operating room and 8^{8,13,38,39,42,45,50,51}  were conducted in the critical care setting. The CNAP device^{10–12,14,15,41} was the most frequently evaluated, followed by T-line,^{16,17,38–40}  Finapres,^50,52–54  Nexfin,^8,9,13  Vasotrac,^43,44  NCAT,^49,51  and others.^{42,45–48,55}  Characteristics of individual studies are presented in table 1.

Results of quality assessment using the modified QUADAS-2 are presented in appendix 4. In all included studies, the risk of bias was assessed as low with regard to patient selection, index test, and reference standard domain. In the flow and timing domains, 22 studies were at low risk. Five studies^{14,42,48,52,55}  were at high risk and one study⁵⁰  was at unclear risk. The concerns regarding three QUADAS-2 applicability domains were all low risk.

Bias and 95% LOA for the 28 included articles are shown in figure 2. Among these, bias and SD for SAP was extractable in 25, DAP in 24, and MAP in 26 studies. Overall, the average continuous noninvasive and invasive SAP were 112.9 ± 19.4 and 118.7 ± 19.4 mmHg, respectively, DAP were 64.3 ± 11.5 and 62.2 ± 10.6 mmHg, and MAP were 78.5 ± 13.1 and 76.9 ± 11.8 mmHg. The overall random-effects pooled bias and SDs of arterial pressure were −1.6 ± 12.2 mmHg (−25.5 to 22.2 mmHg) for SAP, 5.3 ± 8.3 mmHg (−11.0 to 21.6 mmHg) for DAP, and 3.2 ± 8.4 mmHg (−13.4 to 19.7 mmHg) for MAP. We found significant between-study heterogeneity for both biases (P < 0.0001, I²  > 73%) and SDs (P < 0.0001, I²  > 81%) for all arterial pressure variables.

The funnel plots constructed for SAP, DAP, and MAP appeared symmetrical and Egger regression test for asymmetry was nonsignificant (P > 0.1; fig. 3).

We did not find significant differences in biases and SDs in the sensitivity and subgroup analyses. There was, however, significant residual heterogeneity after performing subgroup and meta-regression analysis based on plausible causes (appendix 5).

A forest plot for the sensitivity analysis based only on currently commercially available technologies is depicted in figure 4. Overall, for these studies, the average continuous noninvasive and invasive SAP were 111.1 ± 19.7 and 114.8 ± 19.1 mmHg, respectively, DAP were 64.1 ± 11.6 and 60.1 ± 10.7 mmHg, and MAP were 77.6 ± 13.2 and 75.0 ± 12.0 mmHg. The overall random-effects pooled bias and SDs of 14 studies^{8–17,38–41} were −1.8 ± 12.4 mmHg (−26.2 to 22.5 mmHg) for SAP, 6.0 ± 8.6 mmHg (−10.9 to 22.9 mmHg) for DAP, and 3.9 ± 8.7 mmHg (−13.1 to 21.0 mmHg) for MAP. There was also significant residual heterogeneity with regard to bias and SDs for SAP (I²  = 79.8 and 90.9%), DAP (I²  = 87.3 and 85.3%), and MAP (I²  = 85.4 and 88.6%) within these currently available technologies.

Two of the commercially available devices are based on the volume clamp method (CNAP and Nexfin) and one is based on arterial tonometry (T-line) for the measurement of continuous noninvasive arterial pressure. We conducted sensitivity analysis by device to assess whether significant heterogeneity exited. CNAP and Nexfin showed residual heterogeneity in SAP, DAP, and MAP whereas T-line did not show residual heterogeneity in MAP and DAP (fig. 4).

The average CNAP and invasive SAP were 111.4 ± 20.2 and 118.0 ± 20.6 mmHg, respectively, DAP were 68.5 ± 13.8 and 64.8 ± 11.2 mmHg, and MAP were 74.0 ± 13.8 and 71.1 ± 11.0 mmHg. The overall random-effect pooled bias and SDs were −1.8 ± 12.8 mmHg (−26.8 to 23.2 mmHg) for SAP, 7.2 ± 8.5 mmHg (−9.5 to 24.0 mmHg) for DAP, and 5.5 ± 9.3 mmHg (−12.7 to 23.6 mmHg) for MAP.

The average T-line and invasive SAP were 112.3 ± 18.9 and 113.6 ± 18.6 mmHg, respectively, DAP were 61.8 ± 11.7 and 58.8 ± 10.9 mmHg, and MAP were 79.2 ± 13.2 and 77.9 ± 12.9 mmHg. The overall random-effects pooled bias and SDs were −0.1 ± 8.4 mmHg (−16.5 to 16.3 mmHg) for SAP, 2.9 ± 6.7 mmHg (−10.2 to 16.0 mmHg) for DAP, and 1.3 ± 5.7 mmHg (−9.8 to 12.4 mmHg) for MAP.

The average Nexfin and invasive SAP were 108.9 ± 20.5 and 113.5 ± 18.4 mmHg, respectively, DAP were 63.0 ± 8.0 and 56.5 ± 9.5 mmHg, and MAP were 78.7 ± 12.5 and 74.2 ± 11.6 mmHg. The overall random-effects pooled bias and SDs were −1.6 ± 8.4 mmHg (−18.1 to 15.0 mmHg) for SAP, 5.1 ± 6.6 mmHg (−7.8 to 18.0 mmHg) for DAP, and 3.5 ± 6.8 mmHg (−9.9 to 16.9 mmHg) for MAP.

This meta-analysis of 28 studies assessing accuracy and precision of continuous noninvasive arterial pressure monitoring systems compared with invasive arterial pressure measurements in the operating room and critical care settings showed that the overall random-effects pooled bias of arterial pressure was −1.6 ± 12.2 mmHg for SAP, 5.3 ± 8.3 mmHg for DAP, and 3.2 ± 8.4 mmHg for MAP. When analysis was limited to commercially available technologies, the overall random-effects pooled bias of 14 studies was −1.8 ± 12.4 mmHg for SAP, 6.0 ± 8.6 mmHg for DAP, and 3.9 ± 8.7 mmHg for MAP. On the basis of these results, these devices would not satisfy the standards of the AAMI guidelines as they were interpreted.⁷ 

The goal of this meta-analysis was to assess the accuracy and precision of continuous noninvasive arterial pressure monitoring systems in order to inform clinicians about what should be expected from these devices in the clinical practice. We have identified 28 studies including a total of 919 patients for inclusion in this meta-analysis. These studies included between 8 and 100 patients, and only two of these studies (Ilies et al.¹⁰  and Hahn et al.¹¹ ) included 85 patients or more as recommended by the AAMI guidelines. However, even though Hahn et al.¹¹  included 100 patients, their study evaluated two versions of the CNAP monitor—software v3.0 and v3.5—and only 50 patients only were studied for each version of the software. This meta-analysis allowed assessment of accuracy and precision on a much larger number of patients, increased the power of the study, and allowed subgroup analysis. Because some of these devices have recently been made commercially available and because arterial pressure management is crucial in the perioperative and critical care setting, it is important to have a clear understanding of the accuracy and precision of these systems before use in a clinical setting.

The goal of the continuous noninvasive arterial pressure monitoring devices is to bridge the gap between noninvasive but intermittent and continuous but invasive arterial pressure measurements. However, based on the results from the current study, healthcare providers should be cautious when using these new technologies. For example, if SAP measured using an invasive radial artery catheter was 100 mmHg, SAP measured using a currently available continuous noninvasive arterial pressure measurement system could range anywhere between 74 and 123 mmHg.

Ten^8–17  of 14 studies included in this meta-analysis and published since 2006 (all focusing on currently commercially available devices) cite the AAMI guidelines to calculate sample size or to define acceptability. Among the 14 remaining studies, 5^{41,44,45,49,51}  cited these guidelines to define acceptability, and 9^{38–40,42,43,46–48,50,52–55} did not use any standards or defined arbitrary allowable difference to define acceptability. Interestingly, although studies included in this meta-analysis defined bias as the instantaneous absolute difference between noninvasive and invasive measurements, the AAMI guidelines allow a wider range of values to represent “zero error” when a new sphygmomanometer is compared with invasive measurements. Consequently, the way bias was reported in the method-comparison studies using Bland-Altman analysis^18,19  would result in a greater mean error and SD than the way these standards recommend. One possible reason may be that authors cited AAMI standards and replicated previously published methodologies without reading the original reference. It has been reported that approximately 80% of authors have not read all the articles they are citing.⁵⁶  This leads to the publication of articles with errors or improper citations, which may eventually be responsible for the propagation of misleading knowledge.

It is of major importance for our community to clearly define what level of performance should be expected from these technologies and how method-comparison studies should be conducted and reported. During the past decade, the anesthesiology and critical care communities have been at the forefront of the evaluation of new noninvasive monitoring technologies such as noninvasive cardiac output and hemoglobin monitoring systems. Interestingly, these two device categories are similar to each other in the way they have been developed and tested. After being cleared by regulatory agencies, they were released in the market and extensively tested by clinical scientists who often published controversial results. The lack of consistency in the conduct and report of these studies raised awareness of the need for better standards for method-comparison studies conducted in the perioperative setting.⁵⁷  The results from the current meta-analysis show a similar lack of consistency in the way continuous noninvasive arterial pressure monitoring systems are tested in clinical studies. Different methodologies have been used to evaluate these systems using different thresholds (some inappropriate) to define “acceptability.” This is concerning considering the importance of arterial pressure management in the perioperative and critical care setting. For this reason, the observed heterogeneity between studies and the lack of consistency in the way acceptability of these devices is defined are concerning and should lead our community to adopt more specific standards for conducting and reporting method-comparison studies.

Although we found significant heterogeneity between studies included in our meta-analysis, we did not identify a cause despite performing a series of sensitivity analyses and meta-regressions. It is possible that multiple factors are responsible for the heterogeneity, with each factor making a small contribution. This has been observed in previously published meta-analysis for method-comparison studies.^58–60  Use of different devices in different populations as well as the quality of the studies included in the meta-analysis may also cause heterogeneity, but our results were not conclusive.

Our meta-analysis only assesses the relative accuracy of continuous noninvasive arterial pressure monitoring systems; it does not assess the potential clinical utility of these devices. Indeed, clinical decision-making, patient outcome, and/or patient safety encompass more than the assessment of accuracy and precision of a device. For instance, despite the relatively weak accuracies and precisions of mini- and noninvasive cardiac output monitoring systems,⁶¹  several studies have found a positive impact of these technologies on postoperative outcome (morbidity and length of stay in the hospital) when they are coupled with a protocol defining hemodynamic management strategies.⁶²  However, we believe that arterial pressure is such an important variable for patient safety that strong recommendations about the way these systems should be evaluated and strong evidence related to their accuracies should be reported before any outcome study is conducted.

Another limitation is the mixture of different devices included in this meta-analysis. In particular, the newer systems (CNAP, T-line, and Nexfin) are based on different technologies.⁴  CNAP and Nexfin are based on the volume clamp method and measure arterial pressure at the finger; CNAP is calibrated on an oscillometric arterial pressure cuff whereas Nexfin is uncalibrated. The T-line, however, is based on arterial tonometry and measures arterial pressure from the radial artery. Despite these differences, however, the subgroup analyses showed very similar bias and precision for these three devices. Moreover, our goal was not to evaluate each device but rather to describe the overall accuracy of these technologies. Interestingly, the results from the sensitivity analysis suggest that there is no significant difference between new devices (studies published from 2006) and older technologies.

Finally, our research strategy was limited to studies in English, to PubMed, Web of Science, Cochrane Library, and to articles provided by manufacturers, and only included studies published in peer-reviewed journals and this may induce a bias. We purposely limited the search to peer-reviewed publications in order to avoid low-quality articles. Recent studies have suggested that the extent and effect of language bias has diminished in recent years because of the shift toward publication of studies in English even in national journals¹⁵  and the impact of the inclusion of “gray literature” in meta-analyses is still unclear, may itself introduce bias, and has not been evaluated for meta-analyses of method-comparison studies. Because we were not able to combine results coming from different statistical approaches, we only included studies that provided bias and SD or LOA. Different search strategies (especially using different languages) may have produced different results.

In conclusion, the results from this pooled, weighted meta-analysis demonstrate that the overall random-effects pooled bias of continuous noninvasive arterial pressure compared with invasive arterial pressure measurements was −1.6 ± 12.2 mmHg for SAP, 5.3 ± 8.3 mmHg for DAP, and 3.2 ± 8.4 mmHg for MAP. When analysis was limited to currently commercially available technologies evaluated since 2006, the overall random-effects pooled bias was −1.8 ± 12.4 mmHg for SAP, 6.0 ± 8.6 mmHg for DAP, and 3.9 ± 8.7 mmHg for MAP. On the basis of these results, these devices would not satisfy the standards of the AAMI guidelines. However, most studies evaluating these devices did not report bias and error the way the AAMI recommended, and following the recommendations in that standard would have led to significantly lower error values. Considering the importance of arterial pressure in the management of patients in the perioperative and critical care settings, this study suggests that there is a need to clearly define how these devices should be evaluated and what should be demonstrated to consider them acceptable for use in the clinical setting.

The authors thank Hao-min Cheng, M.D., Ph.D. (Department of Medical Education, Taipei Veterans General Hospital, Department of Medicine, Department of Public Health, National Yang-Ming University, Taipei, Taiwan; Division of Cardiology, Taipei Veterans General Hospital; Faculty of Medicine, National Yang-Ming University, Taiwan), for the kind supply of statistical software for the calculation of pooled data, and Linda Suk-Ling Murphy, M.L.I.S. (Ayala Science Library Reference Department, University of California, Irvine, Orange, California), for the development of search strategies.

Support was provided solely from institutional and/or departmental sources. Dr. Kim was supported from Soonchunhyang University Research Fund (Sinchang-myeon, Asan, South Korea).

Dr. Cannesson is a consultant for Edwards Lifesciences (Irvine, California), Covidien (Boulder, Colorado), Masimo Corp. (Irvine, California), ConMed (Irvine, California), Philips Medical System (Suresnes, France), and Fresenius Kabi (Sèvres, France). A Nexfin monitor (BMEYE B.V., Amsterdam, The Netherlands) and a CNAP monitor (CNSystems, Graz, Austria) were loaned to Dr. Cannesson and his research team in 2010. Dr. Cannesson publicly endorsed the Nexfin technology in a BMEYE newsletter. The other authors declare no competing interests.

1. Blood pressure OR arterial pressure

2. Monitor OR monitors OR measure OR measuring OR measurement OR determinants OR determinant OR determined OR determination

3. 1 AND 2

4. Noninvasive OR non-invasive OR “non invasive”

5. 3 AND 4

6. Nexfin OR CNAP OR Finapres OR Tensys OR T-line OR TL-200 OR Penaz OR Vasotrac OR volume clamp OR applanation tonometry OR finger cuff OR “pulse transit time“ OR finger OR Wesseling OR vascular unloading

7. 5 OR 6

8. Continuous OR continued OR continual OR continually OR continuing

9. Beat-to-beat OR real time OR real-time OR simultaneous OR simultaneously

10. 8 OR 9

11. 7 AND 10

12. Preoperative OR pre-operative OR peri-operative OR perioperative OR intra-operative OR intraoperative OR post-operative OR postoperative OR anesthesia OR anaesthesia OR anesthesiology OR anaesthesiology

13. Surgery OR surgical OR operation OR operative OR operating

14. Critical care OR intensive care OR ICU

15. 12 OR 13 OR 14

16. 11 AND 15

17. Accuracy OR precision OR reliability OR validity OR validation OR standard deviation

18. Bias OR mean difference OR limit of agreement OR Bland Altman

19. 17 OR 18

20. 16 AND 19

Colquhoun DA, Forkin KT, Dunn LK, Bogdonoff DL, Durieux ME, Thiele RH: Non-invasive, minute-to-minute estimates of systemic arterial pressure and pulse pressure variation using radial artery tonometry. J Med Eng Technol 2013; 37:197–202

Reason: Insufficient or lack of demographic data.

Jagadeesh AM, Singh NG, Mahankali S: A comparison of a continuous noninvasive arterial pressure (CNAP) monitor with an invasive arterial blood pressure monitor in the cardiac surgical ICU. Ann Card Anaesth 2012; 15:180–4

Reason: Patient under 18 yr of age (age >16 yr of age).

Stover JF, Stocker R, Lenherr R, Neff TA, Cottini SR, Zoller B, Béchir M: Noninvasive cardiac output and blood pressure monitoring cannot replace an invasive monitoring system in critically ill patients. BMC Anesthesiol 2009; 9:6

Reason: Insufficient or lack of demographic data (could not determine whether bias was calculated as noninvasive minus invasive or vice versa) and retrospective study.

Steiner LA, Johnston AJ, Salvador R, Czosnyka M, Menon DK: Validation of a tonometric noninvasive arterial blood pressure monitor in the intensive care setting. Anaesthesia 2003; 58:448–54

Reason: Insufficient or lack of demographic data (data were expressed as bias and 10th and 90th percentile. No SD could be calculated. Also, there was no extractable age and sex information).

Awad AA, Ghobashy MA, Stout RG, Silverman DG, Shelley KH: How does the plethysmogram derived from the pulse oximeter relate to arterial blood pressure in coronary artery bypass graft patients? Anesth Analg 2001; 93:1466–71

Reason: This study used an ordinary pulse oximeter rather than a device developed for blood pressure monitoring.

Belani KG, Buckley JJ, Poliac MO: Accuracy of radial artery blood pressure determination with the Vasotrac. Can J Anaesth 1999; 46:488–96

Reason: Not in perioperative or intensive care setting (volunteer study).

De Jong JR, Ros HH, De Lange JJ: Noninvasive continuous blood pressure measurement during anaesthesia: A clinical evaluation of a method commonly used in measuring devices. Int J Clin Monit Comput 1995; 12:1–10

Reason: Patient under 18 yr of age (one patient was 17 yr of age).

Wilkes MP, Bennett A, Hall P, Lewis M, Clutton-Brock TH: Comparison of invasive and noninvasive measurement of continuous arterial pressure using the Finapres in patients undergoing spinal anaesthesia for lower segment caesarean section. Br J Anaesth 1994; 73:738–43

Reason: Insufficient or lack of demographic data (the differences between Finapres and invasive systolic, diastolic and mean pressures were considered as a percentage of invasive arterial pressure ((Finapres pressure minus invasive pressure)/invasive pressure × 100)).

Jones RD, Brown AG, Roulson CJ, Smith ID, Chan SC: The upgraded Finapres 2300e. A clinical evaluation of a continuous noninvasive blood pressure monitor. Anaesthesia 1992; 47:701–5

Reason: Patient under 18 yr of age.

Stokes DN, Clutton-Brock T, Patil C, Thompson JM, Hutton P: Comparison of invasive and non-invasive measurements of continuous arterial pressure using the Finapres. Br J Anaesth 1991; 67:26–35

Reason: Insufficient or lack of demographic data (No bias and SD were presented).

Pace NL, East TD: Simultaneous comparison of intraarterial, oscillometric, and finapres monitoring during anesthesia. Anesth Analg 1991; 73:213–20

Reason: Insufficient or lack of demographic data.

Kemmotsu O, Ueda M, Otsuka H, Yamamura T, Winter DC, Eckerle JS: Arterial tonometry for noninvasive, continuous blood pressure monitoring during anesthesia. Anesthesiology 1991; 75:333–40

Reason: Patient under 18 yr of age.

Epstein RH, Bartkowski RR, Huffnagle S: Continuous noninvasive finger blood pressure during controlled hypotension. A comparison with intraarterial pressure. Anesthesiology 1991; 75:796–803

Reason: Patient under 18 yr of age (one 13 yr of aged and one 17 yr of aged males were included).